The arena · AI coding assistant review
Aider
by Aider AI (open source, communautaire)
The free Apache 2.0 CLI pair programmer: 44K GitHub stars, 6.8M installs and roughly 15B tokens flowing through it every week
Free
The tool itself is 100% free (Apache 2.0, pip install aider-chat); the only cost is your own LLM API tokens, typically $30-80/mo for heavy daily use depending on model. Verified against aider.chat and the GitHub repository 2026-07.
Aider AI (open source)
Entire product is free; you only pay your own LLM API tokens
CLI
Yes
Yes
Yes
What is Aider?
Aider is the open source AI pair programmer that lives in your terminal, created by Paul Gauthier in 2023 and licensed Apache 2.0. It edits files directly in your local git repo, auto-commits with sensible messages, and works with almost any LLM (Claude, GPT, DeepSeek, Gemini, local Ollama models) via your own API keys. It pioneered the repo-map technique for large codebases and maintains a public polyglot code-editing benchmark that the whole industry cites.
Aider pros & cons
Pros
- Completely free and Apache 2.0: 44K GitHub stars, 6.8M PyPI installs and ~15B tokens/week routed through it as of 2026; you can audit, fork or self-host everything
- Bring-your-own-model: swap Claude, GPT, DeepSeek, Gemini or local Ollama models per task, so you pay raw API prices (typical heavy use runs $30-80/mo in tokens) with zero vendor lock-in
- Best-in-class git discipline: every AI edit lands as a clean auto-commit you can diff or revert, a workflow even paid rivals copied
- Repo map keeps it effective on large codebases and 100+ languages without uploading your code to a third-party backend
- Battle-tested loop: built-in lint and test hooks let it fix its own errors; 88% of the code in its latest release was written by Aider itself
- Its public polyglot benchmark leaderboard is a de facto industry reference for comparing coding LLMs
Cons
- Terminal-only, no GUI: the learning curve (model flags, API keys, config) deters casual users; watch mode is the only IDE bridge
- Development has slowed: v0.86.2 shipped 2026-02-12 and the cadence dropped from several releases a month in early 2025 to occasional maintenance releases, while Claude Code ships near-daily
- It is a supervised pair programmer, not an autonomous agent: no background tasks, no parallel sessions, no cloud runner; you drive every step
- API costs are your problem: no spend caps or pooled quota, a careless session on a frontier model can burn dollars fast
The arena’s verdict
Take Aider if you live in the terminal, want full control over which model you pay for, and refuse vendor lock-in: nothing else gives you an Apache 2.0 codebase, git-native edits and raw API pricing. Skip it if you want an autonomous agent that runs tasks in the background or a polished IDE experience; Claude Code and Devin are far ahead there, and Aider's slowed 2026 release cadence means the gap is widening. It remains the best zero-dollar entry into serious AI coding, with your only bill being $30-80/mo of tokens for heavy use.
Thumbs up or thumbs down
Cast your verdict
Would you recommend Aider, or warn the crowd away?
Top Aider alternatives
All alternativesCompare Aider head-to-head
What the crowd says
“Feels like it is coasting. Releases slowed to a crawl while Claude Code ships daily. I switched for anything agentic, only come back to aider for quick single-file edits.”
“The auto-commits are the killer feature nobody talks about. Every AI change is a clean git commit I can revert. Saved me twice this week.”
“Still my daily driver after 2 years. Point it at DeepSeek for cheap edits, swap to Opus for the hard stuff. My token bill last month: $41. No subscription can beat that.”
Aider: frequently asked questions
Is Aider free?
Yes, Aider is free to use.
How much does Aider cost?
Aider is free. The tool itself is 100% free (Apache 2.0, pip install aider-chat); the only cost is your own LLM API tokens, typically $30-80/mo for heavy daily use depending on model. Verified against aider.chat and the GitHub repository 2026-07.